For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo module

mha_operand

KV-cache operand descriptors and TMA tile builders for MHA GPU kernels.

Provides the types and helper functions that describe how Q, K, and V tensors (including paged KV caches) are laid out in GPU memory and how TMA (tensor memory accelerator) descriptors are constructed for them.

Structs

• ​KVCacheMHAOperand: An implementation for mo.opaque KVCacheT arguments to MHA kernels.
• ​KVCacheScalesMHAOperand: An MHAOperand that accesses the scales field of a KVCache.
• ​LayoutTensorMHAOperand: An implementation for contiguous tensor arguments to MHA kernels.
• ​RaggedMHAOperand: An implementation for ragged contiguous tensor arguments to MHA kernels.

Traits

• ​MHAOperand: This serves as the trait to support arguments to our MHA kernel.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit